Welcome back to the journal review. This is the April twenty twenty-six issue of JAMA Dermatology, and we've got three pieces worth your time — a full original cohort study on what actually predicts recurrence in early-stage melanoma, a large Danish registry study mapping two decades of keratinocyte carcinoma trends, and a brief economic report doing the unglamorous but genuinely useful math on what sunscreen actually costs your patients. Let's get into it.

First up is an original investigation out of UCSF, "Clinical and Pathologic Factors in Stage One and Two Melanoma Recurrence." The background here is one you already live with clinically: incidence of melanoma keeps climbing, thin melanomas have gotten disproportionately more common relative to thick ones, and yet thin, localized tumors still account for an outsized share of melanoma deaths — in fact, T1 tumors cause more deaths in aggregate than T4 tumors among stage one and two patients, simply because there are so many more of them. Current AJCC staging is built on thickness, ulceration, nodal and metastatic status, and it's genuinely good for prognosticating survival. But the authors point out a real gap: survival prognostication is not the same question as "what predicts recurrence," and recurrence is what patients actually worry about and what drives your surveillance intervals. Prior literature on this is surprisingly thin — pun intended — because a lot of it either lumped all stages together, or restricted itself to patients who underwent sentinel lymph node biopsy, which by current NCCN and ASCO guidelines excludes thin, non-ulcerated stage IA melanomas entirely. So earlier recurrence models systematically underrepresent exactly the population that's numerically responsible for the most deaths.

Methodologically, this is a retrospective single-institution cohort — they pulled every stage IA through IIC melanoma diagnosed between 2010 and 2017 from UCSF's cancer registry, ending up with just over eleven hundred tumors in about a thousand patients after excluding prior recurrences, metastatic-at-diagnosis cases, and duplicates. A retrospective registry design makes sense here for a reason the authors don't spell out explicitly but is worth naming: recurrence is a rare, delayed event, so you need years of follow-up and a large enough denominator that only an established registry can realistically provide — you're not going to prospectively enroll and wait a decade. They used Cox proportional hazards models, first univariable for every candidate variable, then built multivariable models through stepwise selection. One methodologic detail worth flagging for anyone reading recurrence literature critically: several pathologic variables — mitoses, neurotropism, lymphovascular invasion, and comorbidity index — had substantial missingness because they're inconsistently reported on pathology reports. Rather than dropping them or imputing, the authors ran them as separate add-on models to the core multivariable model, one at a time. That's a pragmatic workaround, but it does mean those particular hazard ratios come from smaller, non-identical subsets rather than one unified model — worth remembering when you weigh their effect sizes.

The results. Median age at diagnosis was sixty, cohort was majority male, and — important caveat up front — overwhelmingly White, which will matter for generalizability. Median follow-up was about seven years, and median time to recurrence, when it happened, was around two years, though notably longer for stage one tumors than stage two — three to three-and-a-half years for IA and IB versus roughly one to two years for the stage two substages. Recurrence rate climbed steadily with stage and then plateaued: about four percent for stage IA, fifteen percent for IB, jumping to twenty-eight percent for IIA, and then thirty-seven percent for both IIB and IIC — so the risk essentially flattens out at the top of the localized spectrum rather than continuing to climb. Overall, about one in six tumors recurred. When they did recur, distant recurrence was actually the most common pattern, followed by regional, then local — a reminder that "local" surveillance alone isn't capturing the dominant failure mode.

The multivariable analysis is the real payoff. Six factors independently predicted time to recurrence. The two you'd expect from staging itself were both there and both strong: ulceration carried roughly a three-and-a-half-fold increased hazard, and thickness added incrementally per millimeter — modest per unit, but it accumulates. Beyond those, anatomic location mattered enormously: tumors on the scalp or neck carried about a three-fold higher hazard than tumors on the arm, and face carried roughly a two-fold higher hazard, both statistically significant. Then three histologic features rounded it out — neurotropism nearly doubled the hazard, lymphovascular invasion carried roughly a two-and-a-half-fold increase, and the presence of mitoses — remember, in any amount, not by rate — carried nearly a four-fold increase. All of these were statistically significant, and given the size of these hazard ratios, this isn't a case of statistical significance without clinical meaning — a four-fold or three-fold hazard is the kind of thing that should change how tightly you follow somebody.

One more result worth flagging: among patients who technically met criteria for sentinel node biopsy but didn't get one — for reasons like age, comorbidities, or simply unclear documentation — the recurrence rate was still nearly thirteen percent. That's a population many prior recurrence studies excluded by design, and this cohort deliberately captured them.

On limitations, the authors are appropriately honest: this is one institution, the population is overwhelmingly White and non-Hispanic, so extrapolation to populations with more acral lentiginous or other subtypes is uncertain. The missingness in pathologic variables I mentioned is a real constraint on precision for those specific factors. And it's retrospective registry data, so there's inherent risk of documentation inconsistency and unmeasured confounding — for instance, why some eligible patients didn't get a sentinel node biopsy isn't fully explained by the data available.

So, practically: this is not practice-changing in the sense of altering AJCC staging or biopsy guidelines — that's not what it's designed to do. But it is genuinely useful for the conversation you have with a patient sitting in front of you with a thin, node-negative melanoma on the scalp, or one with incidental mitoses or neurotropism on the pathology report. Location on the head and neck, and the presence of mitoses, neurotropism, or lymphovascular invasion, are independent signals that this particular stage IA or IB melanoma may behave more like a higher-stage tumor for recurrence purposes, even though it isn't upstaged by them. That's a legitimate basis for individualizing surveillance intensity and counseling, even without a guideline change behind it yet.

Second article: a nationwide Danish registry study, "Incidence Trends of Cutaneous Squamous Cell Carcinoma, Carcinoma In Situ, and Keratoacanthoma by Sex, Age, and Anatomical Site." This is a population-based cohort study, and the motivating gap is straightforward — most countries, including the US through SEER, don't systematically register squamous cell carcinoma, carcinoma in situ, or keratoacanthoma the way they register melanoma or internal malignancies, so most of what we know about incidence trends comes from single-center cohorts, insurance claims, or surveys, none of which give a clean national picture. Denmark's registries are unusual in that pathology reporting has been mandatory since the late 1990s, giving essentially complete case capture — which is the authors' own stated rationale for why Denmark is a good laboratory for this question, not something I'm inferring.

The design pulled every histologically confirmed first-time diagnosis of squamous cell carcinoma, carcinoma in situ, and keratoacanthoma in adults from the Danish Pathology and Cancer Registries between 2005 and 2023, using standardized topography and morphology codes, and calculated age-standardized incidence rates plus estimated annual percentage change over the full period. They used Poisson regression for age-specific rates and, for carcinoma in situ specifically, a spline model with a breakpoint in 2018, because the trend visibly changes shape around then.

The numbers: almost a hundred and ten thousand histologically confirmed cases across roughly ninety-five thousand people. Squamous cell carcinoma incidence rose steadily in both sexes — men's rates increased by about two and a half percent per year, women's by about three percent per year — and by 2023 the absolute rates were around one hundred thirty per hundred thousand for men versus about seventy-eight per hundred thousand for women. So men still have the higher absolute burden, but the female rate is climbing faster proportionally. Carcinoma in situ told a more dramatic story — a steep acceleration after 2018, with both sexes converging on nearly identical rates by 2023, in the high eighties per hundred thousand, an annual increase in the range of six percent. Keratoacanthoma, by contrast, actually declined over the study period — though the authors flag that the recent World Health Organization reclassification of keratoacanthoma as a distinct entity may be partly responsible for some of that shift rather than a true drop in underlying disease.

The anatomic and demographic patterns are the part I'd actually use clinically. Men had roughly double the proportion of scalp and neck lesions compared with women, and face was also male-predominant. Women, conversely, had roughly double the burden on the lower limbs across all three diagnoses. And here's the finding I'd flag as most clinically interesting: women aged forty to fifty-nine actually had higher incidence rates than men of the same age, across squamous cell carcinoma, carcinoma in situ, and keratoacanthoma alike — a crossover that only reverses at older ages, where men pull ahead again, particularly past sixty. Reassuringly, trends in people under fifty were essentially flat for squamous cell carcinoma and keratoacanthoma, which the authors interpret — appropriately cautiously — as a possible signal of early prevention efforts finally showing up in the data, though this is an ecological registry finding, not a causal test of any specific intervention.

Limitations here are the ones inherent to any registry study: only histologically confirmed cases are counted, so anything treated clinically without biopsy — which is common for superficial or classic-appearing lesions in some practice settings — isn't captured, meaning true incidence is likely underestimated across the board. And a Danish, largely homogeneous population may not generalize cleanly to more racially and ethnically diverse settings, though the biology of UV-driven carcinogenesis is not exactly Denmark-specific.

Practically, this doesn't change how you treat a squamous cell carcinoma sitting in front of you, but it should sharpen your index of suspicion by demographic: don't let a lower-extremity lesion in a woman in her forties or fifties get dismissed the way you might on an arm, because that's exactly the site and age group where the data show a disproportionate and rising burden. And more broadly, this is useful ammunition for anyone arguing for more capacity and more systematic tracking of keratinocyte carcinoma in the US — the burden is real, it's rising, particularly for carcinoma in situ, and sex-based site patterns are shifting the classic "old man's disease on the head" mental model.

Last one is a brief report, not a full study with a clinical hypothesis being tested — "Sunscreen Costs in Association With Sun Protective Behaviors," an economic evaluation out of the same UCSF group. The motivating problem is one you've probably counseled around without ever quantifying: sunscreen reduces skin cancer and photoaging, adherence is poor, and price is a known driver of underuse and underapplication. Prior cost estimates were from 2011 and 2014, both stale, and used less precise body surface area methods.

Methodologically, this is a straightforward cost-modeling exercise rather than a patient-level study, which is exactly the right tool for this question — they didn't need human subjects, they needed arithmetic done well. They selected three SPF fifty lotion sunscreens with nearly identical active ingredient profiles at low, median, and high price points from an online retailer, then used a modified Lund-Browder burn chart — the same tool you'd use to estimate burn surface area — to calculate exactly how much sunscreen is needed to cover specific combinations of exposed skin at the recommended two milligrams per square centimeter thickness, under different clothing scenarios and usage patterns like a beach week or a full year of an indoor versus outdoor job.

The results are the kind of numbers worth quoting directly to a patient. Unit price for functionally equivalent SPF fifty sunscreens varied by seventeen and a half fold, from under sixty cents an ounce to ten dollars an ounce. Full-body coverage, palms and soles excluded, requires about thirty-three milliliters — roughly one shot glass — per application. A single application's cost ranged from about four cents up to nearly four dollars depending on how much skin clothing already covered and which sunscreen you bought. Stack the cheapest sunscreen with maximal clothing coverage against the priciest sunscreen with minimal clothing, and you get a hundred-and-five-fold spread for one application. Annualized, a year of appropriate sunscreen use ranged from about forty dollars to over fourteen hundred dollars — a thirty-six-fold difference — depending entirely on unit price and how much protective clothing was worn alongside it. Even something as specific as a week at the beach ranged from under seven dollars to over a hundred and thirty-five dollars depending on whether someone wore shorts versus a bikini and which product they bought.

The discussion connects this to a real behavioral mismatch: survey data show consumers say they'd pay around thirty dollars a month for an ideal sunscreen, yet most people report spending less than fifty dollars in an entire year — meaning the willingness-to-pay ceiling and actual spending don't line up, which likely reflects underapplication rather than restraint. The authors' practical suggestion is to frame a full-body application as "about a shot glass" for patient education, and to explicitly counsel that pairing lower-cost sunscreen with hats and long sleeves is not a downgrade in protection, it's a legitimate way to reduce both surface area needing coverage and total cost — which may improve real-world adherence.

Limitations are honestly stated: this all assumes average body surface area and idealized application, doesn't know what people actually pay at checkout, excludes the cost of the clothing itself, and doesn't account for people who might reasonably use a pricier product on the face and a cheaper one on the body. So don't treat these dollar figures as anything more precise than illustrative modeling.

Is this practice-changing? Not in the sense of new evidence linking cost directly to adherence outcomes — that data doesn't exist here. But it's immediately actionable as a counseling tool: you now have concrete numbers to hand a patient who says sunscreen is too expensive, and a clear, evidence-grounded rationale for recommending broad-brimmed hats and long sleeves as a cost-reduction strategy alongside sunscreen rather than instead of it.

That's the April issue in three papers: a UCSF cohort refining how we think about recurrence risk beyond thickness and ulceration in early melanoma, a Danish registry redrawing the demographic and anatomic map of keratinocyte carcinoma, and a cost-modeling brief report giving you real numbers for the sunscreen-affordability conversation. Thanks for listening, and I'll see you next month.